Hacking our collective sense of reality
Trevor Paglen: How to See Like a Machine: Images After AI
Review by Nick Hackworth
“Technology is an event that has befallen man". Philosopher John Gray's observation is proving truer by the day. Both the multiple threats of artificial intelligence (AI) and the mystery of its fundamental operations should extinguish any lingering delusions about humanity's unique genius and omnipotence. Silicon Valley's ruthless right-wing turn underlines that technology will always be employed to serve the contingent expediency of human ends. Only a few decades ago there was an influential belief that Information Technologies were inherently "technologies of freedom", bound to empower freedom of speech and personal autonomy. Who believes that now?
The artist Trevor Paglen's timely book is set against this epic, dystopic landscape and collects his writings from the last decade on how two technological revolutions, the rise of computer vision and of generative media, have transformed visual culture beyond recognition. Paglen's writing is of a piece with his investigative and research-driven art practice: personal, rigorous and imaginative. Despite the constant stream of think pieces on AI, few are asking the acute questions nor making the revelatory historical and critical connections that animate these essays.
The importance of the central tenets of Paglen's thesis cannot be overstated: that the advent of this new tech-animated image culture is more consequential than the discovery of perspective or photography and the birth of mass consumer culture; that digital images are no longer representational but are now operational interfaces servicing corporate or state interests to intervene ever more insidiously, ubiquitously and granularly in our daily lives to extract value and exert influence; that our traditional theories and language for understanding visual culture are obsolete and need radical updating; and that the true effects, scale and speed of these changes on our societies are being barely grasped. If any of that sounds wildly exaggerated, then I can report that by the end of the book, it all feels grimly irrefutable. In this tech-image world the relationship between images and power is fundamentally different from before, and far more terrifying. Adapting to this new reality, we should be asking not what are images saying about the world, but what they are doing to us and the world.
Instruments of power
Paglen explains all the relevant technology clearly and engagingly. "Computer vision" describes the technologies that machines use to translate the visual world into vectors and mathematical abstractions and create essentially machine-readable images. Despite achieving mass adoption only in the 2010s these technologies are automating a vast and growing amount of the "seeing" and reading of images, creating an invisible, machine-to-machine image world that already dwarfs our own. It encompasses processes such as facial recognition, license plate reading, the monitoring of commercial and industrial spaces and operations. These systems are "extraordinarily intimate instruments of power" that allow penetration into our lives at both far smaller and larger scales than possible before and leave humans mostly out of the analytical loops. Equally disturbing, many of the foundational elements of these technologies, such as the data sets on which computer vision and other systems are trained to understand the human world and inform their analytic operations, are encoded with the often crude assumptions of the scientists and researchers that structured them. Paglen interrogates ImageNet, among the earliest and most influential image training sets, to reveal its horrifically negligent, casual and pervasively discriminatory visual categorisation of the world, full of racist, gender and other biases that now infect the digital systems that increasingly shape our world.
Generative media is the automation of image-making and writing and so on. It has shattered what Paglen describes as the "fragile, fuzzy truce" between perception and reality that characterised the modern, "empirical age". Beyond the existential implications for the cultural field, it is the brutal political and economical repercussions that matter most.
In the most speculative and entertaining essay, Paglen dives into the mind-blowing recent history of the scientific research into human perception and its manipulation that informs some applications of these visual technologies. The story encompasses, inter alia, appalling experiments on kittens, the CIA and magicians and magick. We are already through the looking glass in what Paglen, in a nod to the theorist Guy Debord, dubs the Society of Psyop. Across digital media, "images" are increasingly heavily engineered prompts or "payloads", that come "wrapped" in cultural content, designed to directly activate target neurological circuits. Against these "cognitive injection attacks", which cumulatively amount to a hack on our collective sense of reality, the only likely safe spaces are entirely outside the system. Good luck finding them. It will be up to us to carve them out as best we can.
Nick Hackworth